Content area
Marine plastic pollution demands solutions that are both operationally realistic and energy-efficient. We present a smart waste-to-fuel system that converts marine plastics at ≤ 300℃ via IoT-instrumented, low-temperature thermocatalytic depolymerization, optimized with machine learning (ML). The architecture couples marine monitoring for feedstock forecasting, sensor-rich process instrumentation for real-time observability, and ML models for (i) predicting short-term supply and (ii) set-point optimization to maximize liquid yield and energy efficiency. A Batam-based prototype (5-10 kg·h⁻¹) targets polyethylene/polypropylene-dominant streams operated at 100-300℃ under inert/vacuum conditions with solid catalysts. We report liquid/wax yield, condensate properties (density, viscosity, pour point), specific energy consumption (kWh·kg⁻¹), overall energy balance, uptime/MTBF, and ML accuracy (MAE/RMSE). Compared to fixed set-points, ML-assisted control improves temperature tracking, reduces specific energy, and stabilizes product quality, while supply forecasts inform scheduling. The results demonstrate a practical pathway to low-temperature waste-to-fuel for coastal municipalities, preserving the benefits of Internet of Things (IoT) observability and data-driven optimization under a conservative thermal envelope.
